eternityspring/shuohao-skills
AI 短剧制作的 skill 集合:拆角色、排大纲、出场景与道具设定、写剧本、切分镜 | Agent skills for AI short-drama production — character bibles, adaptation outlines, art bibles, screenplays, storyboards. Runs in Claude Code & codex.
What it solves
This project provides a set of agent-agnostic "skills" designed to automate the complex pipeline of turning a novel into production-ready material for AI-generated short dramas. It bridges the gap between a raw story and the technical assets needed for AI video production, such as character bibles, screenplays, and storyboards.
How it works
Each "skill" is a self-contained directory containing a workflow definition (SKILL.md) and Node.js scripts. These skills are designed to be used by AI agents (like Claude Code or Codex) that can read these instructions and execute the bundled scripts. The pipeline follows a specific sequence:
- Novel Outline: Adapts the story into a structured outline with cast and beats.
- Character Bible: Creates detailed profiles and design prompts.
- Character References: Generates consistent visual reference images (anchors) for characters.
- Art Bible: Establishes visual consistency for scenes and props.
- Screenplay: Writes the script with timing estimates and voice prompts for TTS.
- Storyboard: Generates keyframes and prompts aligned with video generation models (e.g., MiniMax H3).
An included assembler script can merge these individual reports into a single, navigable HTML page.
Who it’s for
It is built for AI video creators and producers of short-form dramas who want to standardize their pre-production workflow using AI agents.
Highlights
- Agent-Agnostic: Works with any agent capable of reading
SKILL.mdand run Node.js scripts. - Zero Dependencies: Skill scripts use only the Node.js standard library.
- Production-Ready: Includes specific tools for timing estimation, voice prompt generation, and video model prompt alignment.
- Visual Consistency: Features a dedicated system for generating character anchors to ensure visual stability across different views.
Related
- Project
- Project
- Project
- Project
- Project